The adoption of a performance-based design approach is gaining momentum among tailings engineers, as recommended by the Global Industry Standard on Tailings Management (GISTM). Performance-based design generally refers to utilising surveillance data throughout the life cycle of the facility to re-calibrate the existing design parameters and forecast future performance.
This paper presents a performance-based approach used to re-calibrate the design parameters of a fine tailings material with high plasticity to inform subsequent closure works. The calibration has utilised the data collected from multiple instruments during a closure capping trial on the surface of the fine tailings, including vibrating wire piezometers (VWPs), settlement plates and topographic survey. A Modified Cam Clay model was developed, incorporating the laboratory testing results from Rowe Cell, oedometers and triaxial tests. The predicted strengths are compared to the results from in situ seismic cone penetration tests and vane shear tests. The calibrated model was then used to further indicate the performance of the fine tailings and the adjacent dam structure under the design and construction loading conditions, using the finite element software, Optum. The calibration process has demonstrated good consistency between the predicted and measured deformation and strength values, which provided confidence in the inputs for the subsequent risk-informed option study of closure design.
The approach presented in this paper demonstrates a feasible and reliable way to undertake the performance-based design method. Using performance-based design in tailings management can result in more cost-effective and sustainable solutions. It allows designers to consider a wider range of options and optimise the design based on site-specific conditions.

The quality and frequency of detailed laboratory test programs to characterise tailings have greatly improved over the past five years. However, one area where limitations in current commercial engineering practice remains is in accounting for below-slope stress conditions – specifically plane strain conditions with a non-vertical principal stress angle. Although commercially available SGI and NGI direct simple shear testing devices offer some features relevant to such conditions, they also suffer a number of limitations, particularly with respect to an inability to measure principal stresses. This paper presents an approach for the use of a Torsional Shear Hollow Cylinder (TSHC) device in high-level tailings commercial laboratory programs on the basis of an examination of typical below- stress conditions, a review from the literature on the effects of a non-vertical principal stress angle and a critical appraisal of testing options that currently exist for commercial laboratory practice. Finally, some examples of the additional insight provided by the TSHC, and the potentially lower values of strengths (compared to triaxial compression) from various stress-paths relevant to static liquefaction failures, are provided for further illustration.

Considering the large particle size of rock fill materials (ie -1 cm to 1 m), conducting laboratory tests is prohibitively expensive and time-consuming. Consequently, current design methods and analysis of rock fill structures rely heavily on empirical correlations developed based on historical large-scale experimental studies. Extensive use of rock fill in embankment dams, waste rock dumps, offshore structures etc, on the one hand, and lack of a clear understanding of its mechanical behaviour, on the other hand, highlight a clear need for further investigation. To seek an alternative approach, many studies have been conducted employing parallel particle size grading techniques in which a scaled-down assembly with parallel particle size distribution (PSD) is analysed. It should be considered that rock fill behaviour is significantly scale dependent which predominantly stems from the effect of particle breakage phenomenon. Hence, developing a robust technique (experimental or numerical) that can assimilate effective parameters, including particle shape and breakage, ie allowing simulation and study of large-scale assemblies with realistic boundaries, is considered necessary. Particle breakage is a fundamental phenomenon in the mechanical behaviour of rock fill which substantially influences shear strength, deformability, porosity etc, at different stress levels. The effect of particle breakage can be understood by rock mineralogy, particle shape, surface roughness, PSD, and confining stresses. In this study, a computationally efficient breakage method based on the Discrete Element Modelling (DEM) technique with the use of a flexible membrane is investigated. The results of some large-scale triaxial tests (100 × 180 cm), along with UCS tests are analysed and employed to improve and validate the developed DEM model for large-scale triaxial testing.

Undrained shear strength ratios (USRs) are commonly adopted within the industry to characterise tailings strength, allowing for straightforward statistical analysis. In situ, the undrained shear strength (𝑠𝑢) is typically assessed using cone penetration test (CPT) data, that is normalised by the inferred effective vertical stress (𝜎’𝑣) at the time of testing to define the USR. A strong understanding of 𝜎’𝑣, and consequently piezometric pore water pressure (PWP) profile, seepage and pore fluid density, is key to developing representative USRs. However, from CPT, 𝑠𝑢 is a function of only corrected cone resistance (qt), total vertical stress (σv) and cone factor (𝑁𝑘𝑡). As such 𝑠𝑢 is estimated with reasonable confidence from the CPT data, independent of the in situ pore water understanding.
This paper presents considerations for defining undrained strengths within tailings from CPT data, discussing different approaches to characterising strength using statistical methods. This paper highlights the sensitivity of USR to varying interpretations of piezometric PWP. It presents an equivalent approach to represent the tailings’ undrained strength using total stresses, that removes the need to understand the PWP profile/regime. The assessment applies a stochastic model fitting process to the total stress approach to develop a statistical representation of the tailings’ strength, such that a single characteristic strength can be adopted in the stability analyses, similar to the USR approach. The process followed to develop a statistical representation of tailings strength using the total stress approach is presented. This procedure is not intended to replace the USR approach but to provide an alternative screening method when uncertainty surrounds the in situ PWP conditions within a tailings deposit.

Site investigations were conducted to obtain soil parameters for stability and deformation assessment for an upstream tailings storage facility. Three material types that control the tailings dam stability were identified during the site investigations, comprising coarse-grained silty/sandy tailings, fine-grained silty/clayey tailings and residual soil. Advanced in situ tests, including Seismic Cone Penetrometer Tests (SCPT) and Vane Shear Tests (VST) were conducted. A comprehensive laboratory testing program was undertaken, which included direct simple shear (DSS) tests and triaxial tests to assess the Critical State Line (CSL).
An integrated approach was adopted to determine soil parameters, takings advantage of both in situ and laboratory test results. The peak undrained shear strengths ratios (su/σ’v) of the fine-grained silty/clayey tailings were determined from the DSS test results. The CPT cone factor (Nkt) was calibrated using in situ su/σ’v strengths of the materials in relatively thin layers. Subsequently, the in situ strengths of all other layers were calculated using the Nkt factor. Due to partial drainage in the coarse-grained tailings, its peak and residual undrained shear strengths were assessed using triaxial and non-standard VST test results.
The yield stress of the residual soil was determined using laboratory tests, Self-boring Pressuremeter (PMT) and FLACTM numerical modelling results. The undrained shear strength of the residual soil was based on the calculated yield stress, PMT undrained shear strength and DSS test results.

This paper covers the estimation of shear strength parameter values for input into slope stability analysis of tailings storage facilities. The paper describes the current industry practice of selecting a single (sometimes arbitrary) conservative value for each parameter and proposes to use Bayesian statistics to obtain a meaningful probabilistic evaluation of the strength envelope based on the available data and knowledge of the parameters. The concept of prediction intervals was used to define the strength parameters with selected levels of confidence for the analysis. The paper describes the advantages of the Bayesian approach, including mainly the quantification of uncertainty, traceability of the results, incorporation of prior knowledge and update with new information. The paper uses data from a real project to show the shear strength parameter estimation results for both methods and undertakes a 2D limit equilibrium slope stability analysis, highlighting the differences in the results obtained. The authors believe that the proposed method is robust and systematic and aligns with the risk approach to tailings management proposed by the Global Industry Standard on Tailings Management (GISTM). The use of the proposed Bayesian approach provides the required inputs to estimate the probability of failure for an embankment, which is considered a preferred outcome when compared with a factor of safety, as it is usually the practice in the industry. A discussion including further work and the potential to use these techniques in other areas of tailings design and operation is also provided.

Ten geotechnical engineers will generally provide ten different estimates of tailings characteristic values, based on the same testing data set. Depending on which method is used to derive characteristic values the estimates may be vastly different. This paper explores selected current guidelines and recommendations from relevant standards, codes of practice and tailings literature and highlights the lack of consistency in the matter. Illustrative examples are provided to demonstrate how data distribution, which is rarely a textbook normal distribution, and interpretation methods may lead to overly conservative, or unsafe characteristic values for geotechnical design and analysis of tailings dams. Recommendations are provided, based on the authors’ experience, for a consistent and robust approach in data clean up, distribution review and development of characteristic values. The proposed workflow considers the material’s variability, its impact on the design, and the type of analysis carried out such as limit equilibrium or finite element/finite difference analysis.